We all think we know how a transit app works. There’s a map which knows where all the stops and stations are. Then there’s timetable and route information stored in a database. Put the two together inside a computer or a smartphone and *poof* the magic happens.
Well, sort of. Mostly. Actually, nah – it’s not really like that if you want a good transit app.
Those ingredients are certainly part of the mix. But we’ve been talking to Jens Unger from the HERE transit team, and it turns out that it’s quite a bit more complicated.
“A couple of years ago, we thought that with accurate maps and lots of timetable information, that we had transit ‘solved’. But it turns out there are big challenges that we’re only now starting to come to terms with,” says Jens.
“It mostly comes down to the fact that the way people use transit isn’t how algorithms typically calculate the way they will, so we have to add back that human element.”
Let’s look at four of these ‘human touch’ complications.
Now leaving from platform nine and three-quarters
“One thing we realised was that the information that’s provided on the walking times between different platforms or transport modes at the same transport hub was really inaccurate,” Jens tells us.
Timetables and abstract computer models tend to have – at best – a dim approximation of the time it takes to move from one mode of transit to another.
If you arrive at a station at 9.30, then a train that leaves at 9.31 is not a good suggestion. But a basic ‘map + timetable’ transit system will think that’s super-efficient.
Getting from a bus stop to a train platform takes time, and little effort has been made to provide precise calculations.
“What we’ve been doing is having real people take these journeys or manually calculate the times required to make the transitions. There’s no other way.”
It’s a long and wearying journey to get all that information. But it’s necessary – and no other global transit information provider is prepared to make those extra steps at present
The human dimension
People don’t think or act the same way as computers. And the rules they actually use to make decisions are ones that computers don’t understand unless you write them into the system.
“A computer program would suggest you would prefer to walk 30 seconds to a bus stop, wait one minute, then take a one minute bus-ride, rather than walk for three minutes to get to a nearby station.”
It makes the suggestion that you wait for the bus because that choice will get you there 16% faster.
“But people aren’t like that – if the station is three minutes away, they’ll walk, mostly.”
The team’s research with public transit users across the world showed that:
- People want a stress-free route more than they want a quick route (in fact, people say they want the quickest route, but when you look at what they actually do, they take the one that’s easiest – even if it’s a bit longer);
- They want to get where they’re going on time, so reliability is key: so we mustn’t try to make people take risks, like two-minute changes between different transit types;
- People really don’t like changes, especially more than one in a single journey, even if it means that journey takes slightly longer. (One reason is transitions are quite time-consuming and stressful – neither of those factors are ones which computers are very good at recognising).
Jens’ team is turning these illogical, totally human preferences into algorithmic rules that strive to make the options we suggest a lot more palatable to human tastes, rather than what mathematics would choose for absolute efficiency.
But the human factor goes deeper. We’re all different and geographical cultural differences are one of the key ways we differ.
What counts as an ‘acceptable’ walking distance varies enormously from country to country, and even city to city within those countries.
Similarly, the ‘acceptability’ of different modes of transport varies dramatically, and computers don’t really know much at all about social conventions and tastes.
Again, we need to go back to the ground: “We are involving all of our international offices to understand more about the cultural preferences of each region,” says Jens.
“The results can be very educational, and cut across stereotypes. For example, everyone thinks that Americans won’t walk anywhere, preferring to always drive a car.”
“In actual fact, what we’ve found is that Americans who use transit are prepared to walk further than average – and cycle.”
Bad, bad data
We get our transit data fresh from the source: the transit authorities that provide the services.
That sounds like a good thing, and mainly, it is. “But the transit authorities – understandably – create the timetable and route information for their own use, rather than for interfacing it with data from other transit carriers in the area. They often don’t fit very easily,” says Jens.
“Buses, as we all know, don’t run according to the timetable – they run according to the traffic.” Some transit authorities provide minute-by-minute updates for buses, but most don’t. As a general rule, though, the bus timetable as printed is a convenient fiction.
Again, getting people to take the actual journeys and looking for and integrating real-time API information whenever it’s available is the only possibly answer. No-one other than HERE is doing this on a global scale.
Work in progress
At HERE, we recognise that transit choices and transit truths are enormously complex. It’s not just a ‘mash-up’ of various databases.
And that it’s more complex than other global transit information suppliers care to recognise.
So, no, you can’t just take a timetable database and a computerised map and stick them together. The real world is far more complex than we realise.
But every day, every week, we’re adding more and more intelligence to our systems. So they match real-world expectations, become better than any other system at its scale. And again, HERE is the only global transit information provider that’s doing this.
So next time your transit journey runs just as it should on HERE, thank Jens and his team. And if it doesn’t, well… we’re on the case.